Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Michael Atiyah wikipedia , lookup
Sheaf (mathematics) wikipedia , lookup
Poincaré conjecture wikipedia , lookup
Brouwer fixed-point theorem wikipedia , lookup
Surface (topology) wikipedia , lookup
Orientability wikipedia , lookup
Continuous function wikipedia , lookup
Fundamental group wikipedia , lookup
Covering space wikipedia , lookup
Grothendieck topology wikipedia , lookup
Metrisability of Manifolds in Terms of Function Spaces David Gauld and Frédéric Mynard July 4, 2001 1 Metrisability of manifolds and topological properties of function spaces It is an obvious and well known fact that several topological properties that are different in general may collapse in the presence of additional properties. These additional properties may be algebraic (e.g., a topological group is metrisable if and only if it is first-countable) or purely topological. For instance, a large collection of topological properties which are different in general turns out to be all equivalent to metrisability for topological manifolds(1 ) [6]. Another class of important topological objects in which several different topological properties may collapse is that of function spaces. Let Cp (X) and Ck (X) denote the set of real-valued continuous functions on a topological space X endowed with the topology of pointwise convergence and with the compact-open topology respectively. In general, Fréchetness implies sequentiality and sequentiality implies k-ness, but none of these implications can be reversed. However, these three properties coincide for function spaces like Cp (X) and Ck (X) [16]. In this paper, we show that (not surprisingly) even more properties collapse for function spaces over a topological manifold. Often more surprising are the new criteria of metrisability of a manifold that we derive, in terms of the topological properties of the function spaces over this manifold. Before we state these new criteria, we need to define the notions involved. Definition 1. A topological space X is 1. metaLindelöf if each open cover has a point-countable open refinement; 2. Cečh-complete if it is a Gδ -subset of a compact space; 3. pseudocomplete provided that it has a sequence hBn i of π-bases (B ⊂ 2X is a π-base if every non-empty open subset T of X contains some member of B) such that if Bn ∈ Bn and Bn+1 ⊂ Bn for each n, then Bn 6= ∅; n∈ω 4. cosmic ([8, page 259]) if it has a countable network, i.e., a countable collection N such that if x ∈ U with U open then x ∈ N ⊂ U for some N ∈ N ; 5. a σ-space if it has a σ-discrete (that is, a countable union of discrete families) network; 6. a (strong) Σ-space if there exists a σ-locally finite (i.e., countable union of locally finite families) family F and a cover C by closed countably compact (compact) sets such that whenever C ∈ C and U is an open set that contains C, there exists F ∈ F such that C ⊂ F ⊂ U; 1 i.e., a connected, Hausdorff space which is locally homeomorphic to euclidian space. 1 7. a q-space if each point admits a sequence of neighbourhoods Qn such that xn ∈ Qn implies that hxn i has cluster points; 8. of point-countable type if each point admits a sequence of neighbourhoods Qn such that every filter that meshes every Qn has cluster points. 9. Fréchet if whenever x ∈ A, there exists a sequence hxn i in A that converges to x; 10. sequential if sequentially closed and closed sets coincide; 11. a k-space if A ⊂ X is closed whenever A ∩ K is closed in K for every compact subset K of X; 12. ω-tight or has countable tightness if whenever x ∈ A, there exists a countable subset B of A such that x ∈ B; T 13. ω-fan-tight or has countable fan tightness ([2]) if whenever x ∈ An , there exists finite n∈ω S sets Bn ∈ An such that x ∈ Bn ; n∈ω 14. an ℵ0 -space ([7, page 493]) provided that it has a countable k-network, i.e. a countable collection N such that if K ⊂ U with K compact and U open then K ⊂ N ⊂ U for some N ∈ N; 15. a ℵ-space ([7, page 493]) provided that it has a σ-locally finite k-network; 16. analytic if it is a continuous image of a Polish space (equivalently of the irrationals with their usual topology); 17. hemicompact [1, page 486] if there is a sequence hKn i of compact subsets such that each compact subset of X is contained in some Kn ; 18. Hurewicz (originally called E* in [10, page 195]) if for each sequence hUn i of open covers there is a sequence hVn i such that ∪n∈ω Vn covers X and Vn is a finite subfamily of Un for each n ∈ ω. Even if we primarily focus on spaces of real-valued continuous functions, we also consider conditions on hyperspace topologies. Let C(X) denote the set of all closed subsets of X. This set is classically identified with the set of continuous functions from X to the Sierpiński topology $ (i.e., topology on {0, 1} with ∅, {0, 1} and {0} as open sets), by identifying closed sets with their characteristic functions. Thus, if C is endowed with the cocompact topology, that admits the sets {F ∈ C(X) : F ∩ K = ∅} for K ranging over every compact subsets of X as a subbase, we denote by Ck (X, $) the resulting topological space. It is proved in [14] that C(X) endowed with the upper Kuratowski convergence is first-countable if and only if it is sequential if and only if it is countably tight if and only if X is hereditarily Lindelöf. But hereditary Lindelöfness is known to be equivalent to metrisability for a manifold [6]. On the other hand, it is wellknown (e.g. [4]) that the upper Kuratowski convergence and the cocompact topology coincide for Hausdorff locally compact topological spaces X, in particular for manifolds. This proves equivalences 1 to 4 in Theorem 2 below and gives the first three criteria of metrisability of a manifold in terms of topological properties of function spaces. Theorem 2. Let M be a manifold. The following are equivalent: 1. M is metrisable; 2 2. Ck (M, $) is first-countable; 3. Ck (M, $) is sequential; 4. Ck (M, $) is ω-tight; 5. Ck (M ) is Polish; 6. Ck (M ) is a pseudocomplete σ-space (equivalently is completely metrisable); 7. Ck (M ) is second countable; 8. Ck (M ) is a q-space (equivalently is metrisable, equivalently contains a dense subset of point-countable type); 9. Ck (M ) is a k-space (equivalently is Fréchet); 10. Ck (M ) has countable tightness; 11. Cp (M ) has countable fan tightness; 12. Cp (M ) has countable tightness; 13. Ck (M ) is a ℵ0 -space; 14. Ck (M ) is cosmic; 15. Cp (M ) is cosmic; 16. Ck (M ) is analytic; 17. Cp (M ) is analytic; 18. Cp (M ) is hereditarily separable; 19. Cp (M ) (equivalently Ck (X)) is separable. The following diagrams outline the connections that are true for general function spaces. In the next section, the dotted arrows are shown to be true when X is a manifold. 3 X σ-compact ... metrisable .... ...6 ... .? ... ... Ck (X) cosmic analytic Cp (X).6 .... . ... . . . 6 ...6 .... ... .. .... . ... . . ? ... ... Ck (X) analytic ........ Ck (X) is an ℵ0 -space ... . . . . ... 6 6 .. ... .... ......................................................................................... Ψ . . ......- C (X) Polishœ... Ck (X) is ω-tight k 3 Ck (X) second countable ? - Ck (X) Ck (X) completely metrisable 6 ? [13] Ck (X) contains a dense Cečh-complete subset ? metrisable * ? Ck (X) Cečh-complete ? ] ? Ck (X) first-countable [16] Ck (X) is Fréchet ` ? Ck (X) is sequential ? Ck (X) of point-countable type [12, Theorem 4.4.2] ? Ck (X) is a q-space o Ck (X) pseudocomplete σ-space ? Ck (X) is a k-space [12, p.72] R Ck (X) contains a dense subset of point-countable type œ [12, Corollary 5.6.2, 5.6.3] - Cp (X) separable X submetrisable ... ... 6 ... ... .? [3] - Cp (X) cosmic - Cp (X) hereditarily separable X cosmic œ ... ...6 ... ... ... ... [3] ? .... .......................... œ Cp (X) ω-tight ... X Lindelöf ... 6 .... [2] ...... .- ∀n ∈ ω, X n Hurewicz œ Cp (X) ω-fan tight To prove the dotted implications in the first diagram, we use: 1. [12, Corollary 5.2.5]. Ck (X) is Polish if and only if X is a hemicompact cosmic k-space. 2. [12, Corollary 4.1.3]. Ck (X) is cosmic if and only if X is an ℵ0 -space. 3. [12, Corollary 4.7.2]. Ck (X) is ω-tight if and only if every open k-cover (2 ) has a countable k-subcover. 2 A k-cover of X is a collection S of subsets of X such that each compact subset of X lies in some member of S. 4 4. [12, Theorem 5.7.5]. If X is a q-space (in particular a manifold), then Ck (X) is analytic if and only if Cp (X) is analytic if and only if X is σ-compact metrisable. To complete the proof of Theorem 2 (and in particular dotted arrows in the two above diagrams), it remains to show that the following are equivalent for a manifold M : 1. M is hemicompact and cosmic [equivalently Ck (X) is Polish, as a manifold is a k-space]; 2. M is metrisable and σ-compact; 3. M is metrisable; 4. every open k-cover of M has a countable k-subcover [equivalently, Ck (X) is ω-tight]; 5. M is an ℵ0 -space [equivalently Ck (X) is cosmic]; 6. M is cosmic [equivalently Cp (X) is cosmic]; 7. M n is Hurewicz, for every n ∈ ω [equivalently Cp (X) is ω-fan-tight]; 8. M is Lindelöf; 9. M is submetrisable [equivalently, Cp (X) is separable, equivalently, Ck (X) is separable]. While proving these equivalences in the next section, we actually prove that even more topological properties that are different in general are all equivalent to metrisability for manifolds. Hence we obtain both internal and external (in terms of function spaces) new criteria for metrisability of manifolds. 2 Internal criteria of metrisability Lemma 3. For a topological space X the following three conditions are equivalent. (a) X is hemicompact; (b) There is an increasing sequence hKn i of compact subsets of X such that each compact subset of X is contained in some Kn ; (c) Every k-cover of X has a countable k-subcover. Proof. (a)⇒(b). If hCn i is a sequence of compacta such that each compact subset of X lies in some Cn , then setting Kn = ∪m≤n Cn gives a sequence satisfying (b). (b)⇒(c). Let S be a k-cover of X and suppose that hKn i is a sequence given by (b). For each n ∈ ω choose Sn ∈ S such that Kn ⊂ Sn . Then {Sn / n ∈ ω} is a countable k-subcover of S. (c)⇒(a). Let K consist of all compact subsets of X. Then K is a k-cover of X so has a countable k-subcover, say {Kn / n ∈ ω}. The sequence hKn i satisfies the definition of hemicompactness. Lemma 4. Suppose that N is a k-network on the locally compact, regular space X. Then N̊ = {N̊ / N ∈ N } is also a k-network for X. 5 Proof. Let K ⊂ U ⊂ X with K compact and U open. Use local compactness and regularity to find a finite collection {C1 , . . . , Cn } of compact subsets of U whose interiors cover K. As ∪ni=1 Ci is a compact subset of U there is N ∈ N such that ∪ni=1 Ci ⊂ N ⊂ U . Then K ⊂ N̊ ⊂ U . Lemma 5. Suppose that U is a collection of open subsets of the locally hereditarily separable space X. If U is point-countable on some dense subset of X then U is point-countable. Proof. Suppose that U is point-countable on the dense subset D ⊂ X. Let x ∈ X and let O be an open hereditarily separable neighbourhood of x. Let E be a countable dense subset of D ∩ O; then E is also dense in O. Choose ϕ : Ux → E, where Ux = {U ∈ U / x ∈ U } so that for each U ∈ Ux we have ϕ(U ) ∈ E ∩ O ∩ U : this is possible as O ∩ U is a non-empty open subset of O so that E ∩ O ∩ U 6= ∅. Note that for each e ∈ E the set ϕ−1 (e) is countable since U is point-countable at e. As E is also countable it follows that Ux = ∪e∈E ϕ−1 (e) is countable. Proposition 6. Suppose that X is a locally hereditarily separable, locally compact, regular space which has a k-network which is point-countable on a dense subset of X. Then X has a point-countable k-network. Proof. let N be a k-network on X which is point-countable on a dense subset D of X. By Lemma 4, N̊ is also a k-network on X. Clearly N̊ is point-countable on D so by Lemma 5, N̊ is point-countable. Theorem 7. All solid arrows in the diagram below represent general implications and the dotted arrows hold provided that X satisfies the extra conditions noted. 6 locally second-countable ......................................................- second-countable ... ... ... ... compact .6 locally ... ..... .... ..?.. ... q hemicompact + ℵ0 -space ℵ0 -space .... hemicompactœ ... .... ... I .... .... .... ? ) .... ? ... every open k-cover has . . . σ-compact .... .... .... a countable k-subcover ... .... / ? .... ˜ locally compact .... .... ℵ-space cosmic star-countable ... .... ... ? 9 ... k-network ? ? σ-space .... Lindelöf Hurewicz . ... ...6 .. ? .... ................................................................. strong Σ-space . connected ... locally ... separable ? ... Σ-space .. ?.. ?Æ point-countable metaLindelöf .6 ........................................................ k-network ..̀ regular and Fréchet ....regular, locally compact . . . . .... locally hereditarily . . . .. separable Ψ ... k-network which is point-countable at each point of a dense subset Proof. σ-compact implies Hurewicz. We have X = ∪n∈ω Kn for a sequence hKn i of compact subsets of X. Suppose given a sequence hUn i of open covers of X. For each n ∈ ω, Un is an open cover of Kn so has a finite subcover, say Vn . Then Vn satisfies the requirements. Every open k-cover has a countable k-subcover implies Lindelöf. Let U be an open cover of X and let Û consist of all open subsets of X which are finite unions of members of U . Each compact subset of X is contained in a finite union of members of U , hence in a single member of Û . Thus Û is an open k-cover of X: let V be a countable k-subcover. Each member of V is a finite union of members of U , so there is a countable subcollection W ⊂ U such that each member of V is a finite union of members of W. Then W is a countable subcover of U . ℵ0 -space implies that every open k-cover has a countable k-subcover. Let U be an open k-cover of X and N be a countable k-network for X. For each N ∈ N for which there is U ∈ U with N ⊂ U choose one such U ; call it UN . Then {UN / N ∈ N } is a countable k-subcover of U . Cosmic implies Lindelöf. Let U be an open cover of X and N be a countable network for X. For each N ∈ N for which there is U ∈ U with N ⊂ U choose one such U ; call it UN . Then {UN / N ∈ N } is a countable subcover of U . Lindelöf implies hemicompact under local compactness. Use local compactness of X to cover X by open sets which are the interiors of compact sets. Because X is Lindelöf we need only countably many of these open sets to cover X. Thus we have compact sets {Cn / n ∈ ω} such that X = ∪n∈ω C̊n . Let Kn = ∪m≤n Cm . It remains 7 to show that each compact subset of X lies in some Kn . Given a compact subset K ⊂ X, as {C̊n / n ∈ ω} is an open cover of K there is a finite subcover and if n is the largest index amongst the members of such a finite subcover then K ⊂ Kn . Second-countable is equivalent to hemicompact ℵ0 -space under local compactness. Assume X is second countable. To obtain a countable k-network, take the family of finite unions of elements of a countable base. On the hand, a second countable space is Lindelöf, hence hemicompact if X is moreover locally compact (by the above proof). To prove the converse (in fact the equivalence), we use a dual proof. By [12, Corollary 4.5.3], X is a hemicompact ℵ0 -space if and only if Ck (X) is second-countable. Moreover, Ck (X) coincides with Cc (X), the set of real-valued continuous functions on X endowed with the continuous convergence, provided that X is locally compact [11, Theorem 3.2]. But Cc (X) is second-countable if and only if X is second-countable [5, Theorem 1]. A regular Fréchet space with a point-countable k-network is metaLindelöf by [9, Proposition 8.6(b)]. By Proposition 6, a regular locally compact and locally hereditarily separable space with a k-network which is point countable at each point of a dense subset has a point-countable k-network. The remaining implications in the diagram follow directly from the definitions. Corollary 8. Let M be a manifold. The following are equivalent: 1. M is metrisable; 2. M is second-countable; 3. M is a hemicompact ℵ0 -space; 4. M is hemicompact; 5. M is an ℵ0 -space; 6. M is cosmic; 7. M is an ℵ-space; 8. M has a star-countable k-network; 9. M has a point-countable k-network; 10. M has a k-network which is point-countable at each point of a dense subset; 11. M is metaLindelöf; 12. M is Lindelöf; 13. M is Hurewicz; 14. M is σ-compact; 15. every open k-cover of M has a countable k-subcover. Proof. A manifold is regular, Fréchet, connected, locally compact, locally hereditarily separable and locally second-countable. Moreover, it is well-known (see for example [6, Theorem 2]) that a manifold is metrisable if and only if it is Lindelöf. 8 In particular, the first eight properties quoted at the end of the first section are equivalent (including the seventh, because if a manifold M is Hurewicz, it is moreover second-countable, so that each finite power is also second-countable, hence Hurewicz). Hence all the properties involved in the above diagram except those in the dotted box are criteria for metrisability of a manifold. What about these three additional properties? Definition 9. A topological space X is a Moore space if it is regular and has a development, i.e., a sequence hUn i of open covers such that for each x ∈ X the collection {st(x, Un ) : n ∈ ω} forms a neighbourhood basis at x. X has a (regular) Gδ -diagonal if its diagonal is a (regular) T Gδ subset of X × X (H is a regular Gδ -set in Y if H = n Un , where each Un is an open set containing H). A space X has a G∗δ -diagonal if there exists a sequence hGn i of open covers such T that for each x ∈ X, {x} = n st(x, Gn ). X is weakly normal provided that for every pair A, B of disjoint closed subsets of X there is a continuous function f from X to a separable metric space such that f (A) ∩ f (B) = ∅. It is well-known that each Moore space is a σ-space [7, Theorem 4.5] and that every σ-space has a G∗δ -diagonal [7, Theorem 4.6]. On the other hand, by [7, Theorem 2.15], a locally compact and locally connected space (in particular a manifold) with a G∗δ -diagonal is a Moore space. Hence Moore spaces, σ-spaces and spaces with a G∗δ -diagonal coincide for manifolds. Moreover, there exists a non metrisable Moore manifold [15, Example 3.7]. Thus, none of the properties in the dotted box is a general criterion for metrisability of a manifold. However, it is known from [6] that “normal Moore space” and “weakly normal space with a G∗δ -diagonal” are two such criteria. Proposition 10. The following are equivalent for a manifold M . 1. M is metrisable; 2. M is a (weakly) normal Moore space; 3. M is a (weakly) normal σ-space; 4. M is a (weakly) normal space with a G∗δ -diagonal; 5. M has a regular Gδ -diagonal; 6. M is submetrisable. Proof. The equivalences between the four first points are obvious from the previous discussion while the equivalence between 1 and 5 follows from [7, Theorem 2.15,b)] that states that a locally compact locally connected space with a regular Gδ -diagonal is metrisable. Finally, a submetrisable space has a regular Gδ -diagonal [7, p. 430]. Question 11. Now we can ask whether normality combined with the weaker property of (strong) Σ-space is also equivalent to metrisability for a manifold. 3 Related notions and remarks. Consider the following stronger variantTof fan-tightness. A topological space X has countable strong fan tightness if whenever x ∈ An , there exist xn ∈ An such that x ∈ {xn : n ∈ ω}. n∈ω In [17], it is shown that Cp (X) has countable strong fan-tightness if and only if X n has the following property (called C 00 ) for every n ∈ ω: for each sequence hUn i of open covers there 9 is a sequence Vn ∈ hUn i such that ∪n∈ω Vn covers X. As C 00 is a slight strenghtening of the Hurewicz property which is equivalent to metrisability for a manifold, property C 00 (and hence countable strong fan-tightness of Cp (M )) is a reasonable candidate to be another criterion of metrisability of a manifold M . However, taking l = 1 in Lemma 12 shows that such a space does not satisfy property C 00 . Hence, the function space Cp (M ) over a metrisable manifold M need not be (and actually is rarely) of countable strong fan-tightness . Lemma 12. Let (X, d) be a nontrivial, connected metric space and l ∈ N. Then there is a sequence hUn i of open covers such that for any sequence hVn i, where Vn ⊂ Un and ∪n∈ω Vn covers X, there is n ∈ ω so that |Vn | > l. Proof. Let µ (X, d)¶and l ∈ N be given. Choose two distinct points a, b ∈ X. For each n ∈ ω let Un = {B x; d(a,b) / x ∈ X}, an open cover of X. Suppose that Un,1 , . . . , Un,l ∈ Un for each 2n+2 l n ∈ ω: we show that {Un,j / n ∈ ω; j = 1, . . . , l} cannot cover X. Suppose to the contrary that {Un,j / n ∈ µ ω; j = 1, .¶. . , l} does cover X. As X is connected there must be x0 , . . . , xk ∈ X such that B xi ; 2d(a,b) ni +2 l = Uni ,ji for some ni ∈ ω and ji = 1, . . . , l with a ∈ Un0 ,j0 , b ∈ Unk ,jk , the pairs (ni , ji ) are distinct for i = 0, . . . , k, and Uni−1 ,ji−1 ∩ Uni ,ji 6= ∅ for each i = 1, . . . , k. Then d(a, b) ≤ d(a, x0 ) + k X d(xi−1 , xi ) + d(xk , b) i=1 ¸ k · d(a, b) d(a, b) d(a, b) X d(a, b) + n +2 + n +2 < n0 +2 + n +2 i−1 i 2 l 2 l 2 l 2 k l i=1 = d(a, b) ≤ d(a, b) k X i=0 k X 1 2ni +1 l 1 2i+1 ¶ µi=0 1 = d(a, b) 1 − k+1 2 < d(a, b). This contradiction proves that if we select at most l members of each Un then those sets selected from ∪n∈ω Un collectively do not cover X. Remark. Note that, if in Theorem 7, we transform the condition that every open k-cover has a countable k-subcover by only requiring that every open k-cover has a countable subcover we obtain a condition equivalent to Lindelöfness. We have some other like failures in the sense that we have conditions which superficially appear to be weaker than the standard condition, and hence may lead to a weaker criterion for metrisability of a manifold, but are not weaker at all. Of course in (1) below we could insist that the subcover be a k-cover, in (2) that the subcovers be lindelöf-subcovers and in (3) that the refinement be a k-cover but then the resulting conditions, while being stronger than the standard conditions in general, will be weaker than, for example, hemicompactness so not really of any interest in the context of manifolds. An ideal open cover is an open cover U such that any finite union of members of U is a member of U and any open subset of a member of U is also a member of U. 10 Failure 13. For any space X the following hold. (1) The following are equivalent: (a) X is Lindelöf; (b) every open k-cover has a countable subcover; (c) every ideal open k-cover of X has a countable k-subcover. (2) The following are equivalent: (a) X is Lindelöf; (b) every open lindelöf-cover has a finite subcover; (c) every open lindelöf-cover has a countable subcover. (3) X is metaLindelöf ⇐⇒ every open k-cover has a point-countable open refinement. Proof of (1)(a)⇐⇒(c). The implication (c)⇒(a) may be obtained from the proof of the non-ideal version of this implication in Theorem 7 above by taking Û to consist of all open subsets of finite unions of members of U . Conversely suppose that U is an ideal open k-cover of X and let V be a countable subcover. Let V̂ be the set of all finite unions of members of V. Then V̂ is still countable, is a subfamily of U and is a k-cover of X. Proof of (3). ⇒ is trivial, so we concentrate on ⇐. Suppose that U is an open cover of X. Then Û, which consists of the unions of all finite subfamilies of U , is an open k-cover of d X. Let W be a point-countable open refinement of Û . For each W ∈ W choose U W ∈ Û d d so that W ⊂ U and then choose U , . . . , U ∈ U so that U = U ∪ . . . U W W,1 W,nW W W,1 W,nW . Let V = {UW,i ∩ W / W ∈ W and i = 1, . . . , nW }. Then V is a point-countable open refinement of U. The example below shows that for general topological spaces, 156⇒14 (hence 156⇒4), 156⇒6 (hence 156⇒5), 136⇒14 and 136⇒6 (hence 136⇒5) in Corollary 8. Example 14. Let X be ω1 with the co-countable topology. Then X is Hurewicz and every open k-cover has a countable k-subcover but X is neither σ-compact nor cosmic. The only compact subsets of X are the finite subsets, so X is not σ-compact. Let hUn i be a sequence of open covers of X. Choose any non-empty U0 ∈ U0 . As X − U0 is countable, for each n = 1, 2, . . . we can choose Un ∈ Un so that {Un / n = 1, 2, . . . } covers X − U0 . Now set Vn = {Un } to get the sequence hVn i as in the definition of Hurewicz. Let U be an open k-cover of X. For each finite subset F ⊂ X choose non-empty UF ∈ U and βF ∈ ω1 such that F ⊂ UF and [βF , ω1 ) ⊂ UF ; the latter follows from co-countability of UF . Define inductively an increasing sequence hαn i as follows. Set α0 = 0. Given αn , there are countably many finite subsets of [0, αn ) so {α / α > βF for each finite F ⊂ [0, αn )} is non-empty; let αn+1 be the least member of this set or αn + 1, whichever is the greater. Let α = limn→∞ αn . It is claimed that {UF / F is a finite subset of [0, α)} is a countable k-subcover of U. Indeed, suppose that F is a finite subset of X and set G = F ∩ [0, α). Then there is n ∈ ω such that G ⊂ [0, αn ). As βG < αn+1 < α it follows that F ⊂ G ∪ [α, ω1 ) ⊂ UG . X is not cosmic for if N is a network for X then for each α ∈ X there is Nα ∈ N such that α ∈ Nα ⊂ [α, ω1 ). If α 6= β then Nα 6= Nβ . Thus N is uncountable. The next example shows that for general topological spaces 56⇒13 (hence 156⇒13, 126⇒13, 56⇒14, 56⇒4, 66⇒13, 66⇒14 and 66⇒4) in Corollary 8. 11 Example 15. The space P of irrational numbers with the usual topology, which is homeomorphic to the product space ω ω see [18, Theorem 3.11], is an ℵ0 -space but is not Hurewicz. P second countable, hence an ℵ0 -space. To show that P is not Hurewicz we look at its homeomorph ω ω . For each function σ : {0, 1, . . . , n} → ω set Uσ = {f ∈ ω ω / f (i) = σ(i) for all i ≤ n}. Let hUn i be a sequence of open covers of ω ω defined by Un = {Uσ / σ = {0, 1, . . . , n} → ω}. Suppose that Vn is a finite subfamily of Un for each n ∈ ω. We show that ∪n∈ω Vn does not cover ω ω . Inductively construct a function f : ω → ω. Let f (0) be chosen so that Uf |{0} ∈ / V0 . If f (i) is defined for each i ≤ n so that Uf |{0,1,...,i} ∈ / Vi then just select f (n + 1) so that Uf |{0,1,...,n+1} ∈ / Vn+1 . It follows that f does not lie in the union of all of the members of ∪n∈ω Vn . References [1] Richard F. Arens, A topology for spaces of transformations, Ann. Math., 47(1946), 480495. [2] A. V. Arkhangel’skiĭ, Hurewicz spaces, analytic sets, and fan tightness of function spaces, Soviet Math. Dokl., 33(1986), 396-399. [3] A. V. Arkhangel’skiĭ, On some topological spaces that arise in functional analysis, Russian Math. surveys, 31: 5(1976), 14-30. [4] S. Dolecki and F. Mynard, Convergence-theoretic mechanisms behind product theorems, Top. and Appl. 104(2000): 67–99. [5] W. A. Feldman, Axioms of countability and the algebra C(X), Pacific J. Math., 47(1973):81–89. [6] David Gauld, Metrisability of Manifolds, http://matu1.math.auckland.ac.nz/∼gauld/research/metrisability.pdf [7] Gary Gruenhage, Generalized metric spaces, in “Handbook of Set-Theoretic Topology” K. Kunen and J.E. Vaughan (eds), Elsevier, 1984, 423-501. [8] Gary Gruenhage, Generalized metric spaces and metrization, in “Recent Progress in General Topology” M. Hušek and J. van Mill (eds), Elsevier, 1992, 239-274. [9] G. Gruenhage, E. Michael, Y. Tanaka, Spaces determined by point-countable covers, Pacific J. Math., 113(1984), 303–312. [10] Witold Hurewicz, Über Folgen stetiger Funktionen, Fund. Math., 9(1927), 193-204. [11] D. C. Kent, K. McKennon, G. Richardson, and M. Schroder, Continuous convergence in C(X), Pacific J. Math., 52(1974):457–465. [12] Robert A. McCoy and Ibula Ntantu, Topological Properties of Spaces of Continuous Functions, Lecture Notes in Mathematics 1315, Springer-Verlag, 1988. [13] Robert A. McCoy and Ibula Ntantu, Completness properties of function spaces, Top. and Appl., 22(1986), 191–206. 12 [14] Frédéric Mynard, First-countability, sequentiality and tightness of the upper Kuratowski convergence, to appear. [15] Peter Nyikos, The Theory of Nonmetrizable Manifolds, in “Handbook of Set-Theoretic Topology” K. Kunen and J.E. Vaughan (eds), Elsevier, 1984, 634–684. [16] E. G. Pytkeev, On the sequentiality of spaces of continuous functions, Communications Moscow Math. Soc. (1982), 190–191. [17] M. Sakai, Property C” and function spaces, Proc. Amer. Math. Soc., 104(1988), 917–919. [18] Jerry E. Vaughan, An Introduction to Images of the Irrational Numbers, Topology Atlas, http://at.yorku.ca/i/a/a/i/16.htm, 1996. 13